Alessandro Pellegata and Vincenzo Memoli Department of Social and Political Sciences, University of Milan

Transcription

1 Can Corruption Erode the Confidence in Political Institutions among European Countries? Comparing the Effects of Different Measures of Perceived Corruption Alessandro Pellegata and Vincenzo Memoli Department of Social and Political Sciences, University of Milan Preliminary draft. Please do not quote without permissions Abstract Corruption, as argued by many scholars, reemerged in the concern of most European countries bringing the awareness that no society is immune to the abuse of power by public officials. At the same time, citizens distrust institutions, are skeptical about parties and generally far from political arena. What are the effects of corruption on the quality of democracy and the level of confidence in political institutions? Scholars that investigate this topic usually employ indicators of corruption developed by aggregating experts judgments. In this paper we also consider indexes based on the perceptions of common citizens. The first part of the empirical analysis explores the levels of corruption perceived by citizens of European countries, stressing the differences with experts judgment. Then taking advantage of several datasets we compare the effects of corruption on the confidence that citizens put in political institutions. Empirical analysis shows that more corrupt countries are characterized by lower levels of democratic performance and less confidence in political institutions. Nevertheless, results obtained also highlight some differences in the effects of several measures of corruption adopted. Paper presented at the XXII World Congress of the International Political Science Association Madrid, Universidad Complutense, July 8-12,

2 Introduction Since the early 1990s, there has been a growing awareness worldwide with regard to the consequences that political corruption has both on economic growth and on public support for democratic institutions. Nevertheless, it is also true that corrupt politicians in mature democracies are typically not punished by voters. Sixty percent of incumbents in the U.S. House of Representatives alleged of or charged with corruption between 1968 and 1990 got reelected (Welch and Hibbing 1997). Studies of other developed democracies, as Italy (Chang et al. 2010), Japan (Reed 1999) Spain (Jimenez and Cainzos 2006) or France (Lafay and Servais 2000) show the same trends. Although in a democracy it should be a truism that citizens distrust corrupt governors and inefficient institutions, this may not holds in reality. In fact, as della Porta (2000) noted, corruption may remain hidden and invisible to the public; it may be tolerated or considered secondary to other more serious problems. Therefore, given this impressive and only apparently paradoxical scenario, which is the relationship between corruption and political support? In the present paper we investigate the impact of corruption on citizens confidence in political institutions among European Union member states. The main hypothesis we advance is that the higher the level of corruption present in a country, the lower the citizens confidence in two key institutions of contemporary representative democracies: the parliament and the government. Corruption is supposed to have a negative impact on institutional confidence because of its detrimental effects on economic performance and the quality of governance. Given the difficulties in measuring the phenomenon of corruption and the serious criticisms faced to the most common indexes gathered from experts surveys as to their reliability, we introduce a new indicator developed based on the perceptions of the general public. Differently from the previous studies, besides the Transparency International s Corruption Perceptions Index and the World Bank s Control of Corruption Index this paper tests a measure of corruption developed aggregating the perceptions of more than 20,000 individuals residents in the twenty seven EU member states. First of all, with a descriptive analysis we compare these two different types of indicators to analyze the analogies and differences between the perceptions of the experts and the common citizens on corruption in European countries. Then, in a multivariate analysis we look at the empirical performance of these indicators assessing if they have the same effects on institutional confidence. What emerges form the empirical analysis that follows is that even though experts and common citizens tend to express similar opinions on the extent to which corruption is diffused in the public sector there are some noticeable differences. Measuring corruption with the perceptions of the general public generates more variability among EU member states. Moreover, in some cases common citizens, compared to experts, over-estimate or under-estimate the problem of corruption in their country. Nevertheless, both experts indexes and the indicators developed based on the perceptions of the general public have the same negative impact on citizens confidence in parliament and government. The inverse relationship between corruption and institutional confidence, irrespective of the measure of corruption adopted, remains significant even controlling for alternative explanatory factors both at the individual and the aggregate level. The paper is structured as follows. In the next section we explain the causal mechanism that links corruption and institutional confidence and we make a brief literature review. The third section discusses the relevance of the changes introduced by this paper compared with previous studies. The fourth section presents a descriptive analysis that compares on one side the two main indexes of political corruption based on experts surveys and, on the other side, the indicator 2

3 developed upon the perceptions of common citizens gathered from the Eurobarometer. The following two sections describes the data and the variables used in the empirical analysis, the methodology applied the main results obtained. Finally, in the last section we discuss the main conclusions drawn from the present research and the implications for further research. Corruption and Political Support: Causal Mechanism and Background Literature The concept of political support assumes particular relevance in comparative institutional analysis because it has several consequences on the functioning and the legitimacy of democratic regimes. David Easton (1965; 1975) distinguished between diffuse support and specific support. Whereas the former refers to the political community and the regime, the latter implies a citizens evaluation of the performance of the current political authorities. Satisfaction on how democracy works in a country and confidence in national government and other political institutions, such as the parliament, the legal system, and civil servants, have been usually considered indicators of specific support (Klingemann 1999). In this study we advance the hypothesis that corruption has a negative impact on citizens confidence in political institutions. Corruption is supposed to generate distrust in political institutions because has detrimental effects on the performance of political systems. Although in the mid 1960s several scholars considered corruption functional to the process of modernization of political systems and to the legitimization of their institutions in front of the citizens (Leff 1964; Huntington 1968), a recent broad literature shows that corruption hinders economic growth (Mauro 1995) and diminishes the quality of governance (Kaufmann et al. 1999). Corruption is defined as the misuse of public office for private financial gain (Rose-Ackerman 1999; Treisman 2000; Warren 2004; Kunicovà and Rose-Ackerman 2005). This definition indicates that, even though there are different activities that can be qualified as corrupted, each of them typically involves elected officials and/or appointed bureaucrats who abuse the power and authority with which they have been entrusted for private gain. Such activities take place at the expense of the collective community, thereby violating the norms that regulate the public office (Warren, 2004; Pellegata 2012). 1 When public agents are committed to deriving personal gain from their privileged positions, human, social, and economic resources are diverted from the public interest undermining the principles of democratic accountability and of fairness and impartiality of political institutions. When high levels of corruption are present institutions become rent-seeking instruments in the hands of political leaders and a small economic elite, while the public is compelled to pay the externalities of this misallocation process consisting in a distortion of public demand, an increase of the cost and a reduction of the quality of public services (della Porta 2000). Corruption can be considered as a source of what the principal-agent theory calls agency loss (see among others Lupia 2003). This term indicates the damage suffered by the principal (the citizen) because an agent (the public official) lacks the skill or incentive to complete the task delegated to her/him. In sum, corrupt institutions mean inefficient institutions that lose their credibility in the eyes of the citizens and, as we can presume, contribute to generate disaffection toward politics and distrust in institutions. 1 For a review of different types of corruption see Rose-Ackerman (1999) and Langseth (2006). 3

4 In the last decade several studies have provided empirical support for a detrimental impact of corruption on political support. These studies present methodological differences, in particular regarding the measurement of corruption, that is worth discussing for the purpose of our research. Mishler and Rose (2001) tested the validity of cultural and institutional theories on the origins of political trust on a sample of ten post communist countries in Eastern and Central Europe and the former Soviet Union. The authors measures the citizens perceptions of corruption through a question taken from the New Democracies Barometer which asks citizens if, in their opinion, the level of corruption decreased, remained the same or increased since Communism. The main findings of this work clearly indicate that individual evaluations of the political performance of countries, including the level of political corruption, are the main determinants of political trust measured with the mean score of individual trust in six institutions, such as parties, parliament, president and/or prime minister, courts, police and military. 2 Chang and Chu (2006) analyzed the effects of corruption on citizens trust in political institutions in five emerging and consolidated Asian democracies (Japan, Philippines, South Korea, Thailand and Taiwan). Relying on East Asian Barometer the two authors measured corruption through the citizens perception on the level of corruption in the national and local governments. Their findings clearly demonstrates a trusteroding effect of corruption even in Asian countries that, according to many commentators, present contextual factors, such as political culture and electoral systems, that should mediate the negative impact of corruption. This result survived also after the authors had taken into account the reciprocal relationship between corruption and trust. Unlike Mishler and Rose (2001) and Chang and Chu (2006), Seligson (2002) measured the level of corruption drawing upon national survey data collected in four Latin American countries (Bolivia, El Salvador, Nicaragua and Paraguay) which assessed not perceptions but rather effective corruption experiences suffered by citizens. Results obtained from the empirical analysis showed that, at least in the four Latin American countries analyzed, corruption is associated with lower levels of support for the legitimacy of political systems, measured through the level of confidence in the key institutions of government. These findings permit Seligson (2002) to refute the Huntinghton s view of corruption as functional for political development of emergent and unconsolidated democracies. Differently from the studies analyzed so far, Anderson and Tverdova (2003) and Wagner et al. (2009) tested the hypothesis that corruption produces lower levels of political support measuring the independent variable at the aggregate country level rather than at the individual level. More precisely, the average level of corruption present in a country is estimated through perceptions of small groups of experts collected in polls and surveys. Anderson and Tverdova (2003) conducted probably one of the most influential study on the effects of corruption on political support on a sample of sixteen contemporary democracies that includes both Western countries and postcommunist countries of Eastern Europe. They hypothesized that the impact of corruption on the degree of satisfaction with democracy and of trust in civil servants expressed by citizens is mediated by people s political allegiances. The authors showed that, irrespective of the statistical technique employed, corruption is negatively correlated with both the dependent variables used. Furthermore, they showed that the trust-eroding effect of corruption is attenuated among supporters of incumbent political authorities. Wagner et al. (2009) conducted a panel analysis on sixteen 2 A similar analysis on a sample of post-communist countries of Eastern and Central Europe using data gathered from New Democracies Eurobarometer is conducted by Wallace and Latcheva (2006). These scholars obtained results in support of a detrimental effect of corruption on citizens confidence in the legal and political system. 4

5 Western European countries in the time span They aimed to analyze longitudinally the effects of the quality of institutions on the citizens levels of satisfaction with the way democracy works in their country taken from Eurobarometer surveys. The quality of institutions is operationalized with different indicators measuring the level of corruption, the regulatory quality, the rule of law and the size of the shadow economy. From the empirical results they obtained emerges that lower corruption, a better rule of law, a smaller shadow economy and a better institutional quality in general are associated with higher levels of satisfaction with democracy. Expanding Previous Research The present research builds upon these scholarly efforts and expands our knowledge of the relationship between corruption and political support in several ways. Firstly, the effects of corruption on institutional confidence are tested through a crosssectional analysis on a sample composed by the twenty seven European Union member states. To our knowledge comprehensive studies on the relationship between corruption and political support among European Union countries do not exist. Analyses conducted on Europe usually focused on subgroups of countries, such as Western European democracies (Criado and Herreros 2007; Wagner et al. 2009; Wallace and Letcheva 2006) or post-communist countries of Central and Eastern Europe (Mishler and Rose 2001). However, we think that European Union constitutes an interesting area for the investigation of the effects of corruption on political support. In fact EU is composed by both mature and newly established democracies and presents a high degree of variation according to both the levels of corruption and of political support. Notoriously, Scandinavian countries are considered among the least corrupted countries in the world and present very high levels of institutional trust. At the same time, some Southern and Eastern European countries tend to present high levels of corruption, comparable to the ones of other less developed regions in the world, and suffer from a general disaffection of citizens toward politics. Secondly, the focus on EU member states assumes further importance because the empirical analysis is conducted relying on individual and country-level data that refers to 2009, when the international economic and financial crisis was already exploded. As we know, this crisis has strongly invested Europe, in particular Greece, Ireland, Italy, Portugal and Spain, and its effects are still present. It is presumable that one of the most evident political consequences of the economic crisis will be a decrease in the citizens confidence toward political institutions. Furthermore, the potential political instability generated by the economic crisis can provide a fertile breeding ground for the abuse of power by bureaucrats aware of the existing institutional inefficiencies. Finally, and most importantly, taking advantage of several datasets, we have simultaneously tested different indicators of political corruption built on the perceptions of both experts and common citizens of the countries included in our sample. All the studies cited above have used either aggregate indicators developed based on expert surveys or the citizens individual perceptions of how corruption is diffuse in their country. In the present paper we adopt a different strategy. We have relied on surveys that collect the perceptions of the general public to build an aggregate indicator that estimates the level of corruption in every country included in our sample. This permit us to assess, through a multivariate analysis, the empirical performance of this new index in explaining the level of political support compared to the most commonly used indicators of corruption based on the perceptions of small group of experts. Of course relying on citizens or experts perceptions have advantages and disadvantages. In the next section we describe how the 5

6 index based on citizens perceptions is developed and we compare this with the most common indicators based on experts perceptions analyzing analogies and differences among them. Measuring Corruption through Perceptions. But Perceptions of Whom? In recent years scholars have experimented different approaches to measure political corruption. 3 As we have seen in the brief literature review presented above, measuring corruption through indicators developed based on perceptions has become the most common approach. In particular, there are two indicators that are usually adopted in comparative analyses of the causes and the consequences of corruption. The Transparency International s Corruption Perceptions Index (CPI) is an annual index that measures how corrupt the public sector is perceived to be. It is released since 1995 and in 2009 covers 180 countries. 4 CPI ranges between 0, indicating highly perceived corruption, and 10, which means low levels of corruption. The Control of Corruption Index (CC) developed by the World Bank measures the extent to which public power is exercised for private gains, including both petty and grand forms of corruption, as well as capture of the state by elites and private interests. It is collected every two years from 1996 to 2002 and every year since 2003 for more than 200 countries and territories. 5 CC ranges from -2.5 (more corruption) to +2.5 (less corruption). Both CPI and CC are developed aggregating a large number of partially similar individual data sources including polls of experts, surveys of businessmen, senior managers and individuals and the assessments of commercial risk agencies, non-governmental organizations, and multilateral aid agencies. 6 Unsurprisingly among EU member states in 2009 CPI and CC are almost perfectly correlated (r=0.9882). Although among the sources of CC there are also some surveys of common citizens, both CPI and CC tend to reflect the perceptions of corruption of the economic elite of different countries and the international private sector. But which is the opinion of the common citizens residents in EU member states on corruption? Do the perceptions of experts reflect the opinion of the general public or tend to be disconnected from it? To answer to this questions we have relied on the Eurobarometer 72.2 released in September 2009 which collects opinions of more than 26,600 citizens of the twenty seven European Union member states on corruption and other issues. More precisely, responses to seven different questions allow us to map the opinions of citizens of EU member states about the extent of corruption at different institutional levels, the experiences with corruption they have suffered, the main reasons of the phenomenon and the efforts that institutions made to constrain it. We have applied a Multiple Correspondence Analysis (MCA) to synthesize the information contained in these indicators which refer to different aspects of the phenomenon. 7 The MCA allowed us to reduce the wide range of variables considered here into a more limited number of factors. Then, using the SPAD N software, we have explored the relationship among the indicators 3 For a review of different approaches used to measure corruption see Seligson (2002) and Sampford et al. (2006). 4 See 5 The Control of Corruption index is one of the six Worldwide Governance Indicators developed by the World Bank. See 6 In 2009 CPI is developed aggregating data gathered from thirteen sources, while CC is built on information coming from thirty individual sources. 7 The active variables, which contribute to the formation of the axes factors, included in our MCA are 90. In this analysis we have not considered the variables which modalities presented a frequency percentage less than 2. 6

7 of the perceived corruption among European countries. The result was a bi-dimensional space with two factors that explain 94.2% of the total variance. 8 Figure 1 here In the graphical representation reported in Figure 1 these two factors correspond to two distinct axes. The first factor (horizontal axis) explains 87.2% of the total variance and the second factor (vertical axis) 7%. The distance of different modalities from the center of the two factorial axes tells us how much they contribute to the definition of the axes. The horizontal axis tells us the perceptions of how corruption is diffuse in different levels of government (local, regional and national level) and with reference to different figures of the public sector (politicians, different bureaucrats). Moving from the left to the right of the horizontal axis means that the level of perceived corruption among citizens increases. The vertical axes refers instead to the responsibility of and the efforts made by the institutions in fighting corruption. European citizens that have a negative view about the problem of corruption are located in the lower-right side of the graph, whereas in the upper-left side are located those citizens who perceive low levels of corruption. As we can note citizens perceptions of corruption at the different levels of governments tend to move together. Citizens who consider corruption as a major problem in their country think that it is widespread both at the national level and in the regional and local institutions. Moreover, these citizens think that corruption is particularly diffuse especially among politicians and the police, as well as in the public health and the business sectors of bureaucracy. Focusing on the vertical axis of the graph, we can note that European citizens think that the prevention from and the fight of corruption is mainly a responsibility of EU institutions and the judicial system. This preliminary analysis of the general public opinions regarding the phenomenon of corruption confirms that there is a high variability among different EU members state. People who live in countries characterized by high levels of civicness and a good performance of public and private institutions, such as Scandinavian and north continental countries (Denmark, Sweden, Finland and Netherlands), perceive very low levels of corruption and are satisfied of how institutions constrain the potential problems generated by this phenomenon. On the contrary, citizens of countries where the economy has low performances and the social capital does not reach high levels, such as Eastern and Southern European countries (Bulgaria, Lithuania and Portugal) consider corruption a major problem, diffused among different institutional sectors. To better compare the general public view of the problem of corruption with the perceptions of experts we develop an index of corruption aggregating the answers citizens have given to three different questions included in the Eurobarometer These questions asked to citizens of EU member states if they agree or disagree with the statements affirming that there is corruption in local, regional and national institutions. 9 Applying a factor analysis to the citizens responses to this questions we obtain a continuous index (factor score) that we call Corruption-Eurobarometer 8 Since eigenvalues obtained by MCA give a pessimistic evaluation of the variability explained by factorial axes, we have computed the variance considering only eigenvalues higher than 1/k, where k is the number of variables used in the analysis. 9 Original answers to these items are four-fold ( totally agree, tend to agree, tend to disagree, totally disagree ). We have dichotomized these answers recoding the first two modalities in agree (0) and the second two categories in the category disagree (1). 7

8 (CEB) that attributes a score to every country included in the sample. As for CPI and CC, the higher the level of CEB, the lower the level of corruption. Figure 2 reports three histograms in which the twenty seven EU member states are arranged according to the corruption score attributed to them by CPI, CC and CEB, respectively. The horizontal black lines in the graphs indicates the value assumed by the sample mean. Figure 2 here According to the perceptions of both the experts and the general public in 2009, Denmark is the less corrupt country, followed, though not in the same order, by Sweden, Netherland, Luxembourg and Finland. In the same way, according to the three indicators, countries of Eastern and Southern Europe, such as Bulgaria, Romania, Lithuania and Greece are placed at the bottom of the ranking. Figure 3 shows two graphs which combine the view of the general public on the extent of corruption in their country, as expressed by CEB with the experts perceptions, reflected in CPI and CC, respectively. As we can see the opinions expressed by common citizens is positively related with the view that emerges from experts surveys: the greater the perceptions of corruption expressed by the citizens of different countries, the greater the perceptions of the experts. More precisely, the Pearson s index of correlation between CEB and CPI is r=0.7756, while between CEB and CC r is equal to This result seems to contradict the claim which affirms that indicators based on experts opinion are disconnected from those of the general public (Kaufmann et al. 1999). At the same time, it also disproves the common wisdom that common citizens, differently from experts, cannot really assess the actual diffusion of corruption in their country s public institutions (Kaufmann and Wei 1999). What emerges from this preliminary analysis is that experts and citizens perceptions tend to move in the same direction. Figure 3 here Nevertheless, in spite of these analogies, there are also some differences between experts and citizens perceptions that is worth discussing. Figure 4 presents three box plots that depict the data distributions according to CPI, CC and CEB, respectively. As we can easily note, whereas the distributions of CPI and CC in our sample are more homogeneous, CEB concentrates almost all the countries at the bottom of the scale. The five less corrupt political systems (Denmark, Netherlands, Luxembourg, Sweden and Austria) instead are outliers in the distribution of CEB. This means that, differently from experts, citizens of these countries perceive levels of corruption that are much more lower than the levels perceived by the citizens of the other twenty two countries included in the sample. Figure 4 here The image depicted in Figure 4 seems to give support to one of the main critique moved toward the use of indexes of corruption developed based on general public perceptions. It is often argued that these indicators are not as reliable for cross-country comparative analysis as the one based on experts perceptions because common citizens tend to make in-context or country-specific interpretations of the phenomenon of corruption (Kaufmann and Wei 1999). In particular it is 8

9 contested the fact that common citizens tend to over-estimate or under-estimate the problem of corruption in their country. An in-depth comparison of the ranking of the sample countries according to the three indexes proposed can help to shed light on this problem. Table 1 here Table 1 ranks the twenty seven EU member states according to the values attributed to them by CPI, CC and CEB respectively. Horizontal black lines delimit the quartiles in which the distributions are divided. Whereas the rankings made on the basis of CPI and CC are almost the same, the ranking of countries made according to CEB presents some significant differences. We have highlighted those countries that, according to the citizens perceptions, change the quartile in which they are placed in respect to experts evaluations. More precisely, in bold are underlined those countries that move from the bottom to the top of the distribution, while in italics we have stressed those that move from the top to bottom of the distribution. Compared to experts evaluations, citizens of Estonia, Italy, Poland and Slovakia for instance seem to under-estimate the problem of corruption in their country. The biggest shift is made by Italy that moves from the first quartile of the CPI and CC distribution to the third quartile of the CEB one. On the contrary, other countries, such as Cyprus, Ireland and Portugal, are judged more negatively by common citizens than by the experts. According to this descriptive analysis we can draw some preliminary evidence on the relationship that links indicators of corruption gathered from experts surveys and indicators developed based on citizens perceptions. Although CPI(CC) and CEB are positively and significantly correlated, indicating that experts and citizens perceptions tend to move in the same direction, this correlation is far from being perfect. This highlights that there are some differences in the way experts and citizens evaluate the problem of corruption in European countries. Firstly, if on one side countries perceived as the least corrupt by the experts and the citizens are the same, on the other the difference between the level of corruption in these countries compared to all the others is much more higher according to citizens evaluations than to the experts ones. Furthermore, there are two apparently contrasting tendencies in act when citizens perceptions are evaluated. On one side citizens under-estimate the situation of some countries (especially Cyprus, Ireland and Portugal) to which experts attributed low or quite low levels of corruption. On the other side, citizens of countries like Estonia, Poland Slovakia and above all Italy, under-estimate the high levels of corruption attributed to them by the experts. With these not-trivial differences in mind we now turn to look at the empirical performance of these indicators in a multivariate statistical analysis that aims to analyze the relationship between corruption and political support. In the next section we will describe all the variables used in the empirical analysis, while in the following one we will explain the statistical method employed and the main results obtained. Data and Variables The dependent variable in our analysis is the citizens confidence in political institutions. More precisely, we analyze the level of confidence expressed by citizens of different European Union member states in the government and the national parliament. These are two key institutions for the functioning of contemporary representative democracies and, consequently, it is presumable that the confidence that citizens posited in them will be highly affected by the spread of corruption among 9

10 politicians and public officials. The indicators selected to operationalize citizens support for these institutions are confidence in parliament (CONFPARL) and government (CONFGOV) included in the Eurobarometer 72.1 of September The answer to these questions is on a scale that goes from 0, no trust at all in parliament (government), to 10, trust completely in parliament (government). Figure 5 shows the average levels of confidence in parliament and government for each sample country. As we can see, in a large part of countries analyzed the mean level of confidence in the parliament tends to coincide with the mean level of confidence in the government. The countries that present the highest levels of support for these two institutions are Denmark, Sweden, Luxembourg and Finland, while the lowest levels of political support are shown by Latvia, Lithuania, Poland and Hungary. Figure 5 here The independent variable is the level of perceived political corruption and is operationalized with the three indexes (CEB, CPI, CC) that we have presented and discussed in the previous section. The higher the level of these three indexes, the lower the level of corruption. The relationship between corruption and confidence in political institutions is controlled for a series of control variables measured at the country level and at the individual level. The country level variables include variables that refer to the political-institutional structure of different countries and variables related to the their economic performance. Constitutional design (MAJ) refers to the countries institutional framework and is operationalized with the index of majoritarianism developed by Lijphart (1999) on the executive-parties dimension. MAJ captures the proportional or majoritarian pattern of democracy present in different countries analyzing several institutional elements, such as the party system, the electoral system, the relationship between the executive and the legislative. More precisely, MAJ is the mean of five indicators the Laakso and Taagepera s (1979) effective number of legislative parties, the Gallagher s (1991) index of voteseats disproportionality, the duration (in days) of the executives, the percentage of single party and minimum winning coalition governments and the index of corporativism measured on a period that goes from the first democratic elections after 1945 to MAJ ranges from 0 to 4, being 0 the lower level of majoritarianism and 4 the higher level. As a consequence, the fifteen countries (among others Belgium, Denmark, Finland, Netherlands) that have a value of MAJ between 0 and 2 are predominantly consensual democracies, whereas the twelve countries (among others France, Ireland, Spain, United Kingdom) comprised between 2 and 4 are predominantly majoritarian systems. 10 As many scholars argued (Lijphart 1999; Powell 2000), in majoritarian systems the power is concentrated in the hands of the incumbent party, whereas in consensual democracies the power tends to be dispersed among different institutions. This implies that in majoritarian systems 10 For every country included in the sample we have computed the mean value for each of the five indicators that compose MAJ for the period that goes from the first democratic election taken since 1945 to Then, following the procedure developed by Lijphart (1999: ), we have standardized these value in order to make them vary on the same scale. The sign of the effective number of legislative parties has been inverted, so that the lower value of all variables indicates the consensual extreme and the higher value the majoritarian extreme. Then, for each country we have computed the mean of the five variables. In this way all countries are displayed on a space determined by two standard deviations above and under the average value of 0. For the sake of simplicity we have transformed the original scale in a scale that ranges between 0 and 4. Data on electoral results, parliaments and governments composition are gathered from ParlGov Dataset ( while data on the index of corporativism are taken from Siaroff (1999) and Roberts (2006). 10

11 citizens can easily attribute responsibility for policy outcomes to the government, whereas in consensual democracies the attribution of the responsibility is much more obscure. Nevertheless, consensual democracies, through proportional representation systems, multiparty systems and coalition governments, favor a higher representativeness of citizens in the parliament and the government. As a consequence, we expect that people who lives in consensual democracies display higher levels of confidence in political institutions that those who live in majoritarian systems. New democracies (NEWDEM) is a dummy variable that assumes value of 1 for those political systems in which democracy has not been in place for at least 25 years, that in our sample coincide with the ten countries of Eastern Europe. As we have previously seen, Eastern European countries tend to present lower levels of institutional confidence and higher levels of corruption than consolidated democracies. There are several factors that lead us think that it is particularly important to control the impact of corruption on confidence in political institutions for the effects of new democracies. In fact, the higher levels of distrust in political institutions in young democracies can be caused, for instance, by rapidly changeable party systems, unstable governments and the legacy that especially older people feel with the past regime (Linz and Stepan 1996, pp ). Moreover, as Horowitz et al. (2009) have shown, the impact of corruption on the quality of democracy and the regime legitimacy is more evident in hybrid regimes and young democracies where a confuse and unstable political context leaves more discretional power to politicians and bureaucrats, which can abuse of their positions to obtain private financial gain. Therefore, we expect to find new democracies negatively associated with citizens confidence in the parliament and the government. The economic performance of the sample countries is operationalized with two variables commonly used in cross-national studies on political support, the level of economic development (loggdp) and the degree of economic growth (GDPG). The first variable is measured with the natural logarithm of GDP per capita in 2009, while the second is given by the average degree of GDP growth in the last five years as a percentage of GDP. 11 According to the findings obtained by past studies, we expect that a better country s economic performance, that means higher levels of per capita income and economic growth, will be associated with higher institutional confidence. At the individual level we have controlled the effects of corruption for the impact of several demographic and socio-economic variables that are usually considered to be correlated with institutional trust. These variables are included in the Eurobarometer 72.1 where we have taken the measures of confidence in the parliament and the government used as dependent variables. Life satisfaction (LIFESAT) is a variable with values that range from 0 for those people very dissatisfied for the quality of their life to 10 for those people that are instead very satisfied. Financial expectation (FINANCEXP) is a dummy variable that assumes value of 1 for those people that think that the financial situation of their household will be better in the next twelve months, and assumes value of 0 for those people that think that their household s financial situation will remain the same or be worse. Social level (SOCLEV) is a variable gathered from an item that asks people to place themselves on a scale that goes from 0 (lowest level) to 10 (highest level) in accordance to their level in the society. We expect that people more satisfied of their life, people trusting that their household s financial situation will be better in the future and people that place themselves in the highest levels in the society will be more confident with the parliament and the government. 11 Data for these two variables are taken from World Development Indicators developed by The World Bank. See 11

12 Moreover, we also include in the analysis other three socio-demographic variables, such as Age, Gender and Education. Age (AGE) simply measures the age of respondents if they are at least 18 years old, while Gender (GENDER) is a dummy variable that assumes value of 1 for women. Education (EDUCATION) measures the age of respondents when they stopped full-time education. Analysis and Results The empirical analysis is conducted on a dataset that combines more than 23,400 observations at the individual level of respondents nested with information at the aggregate level of the twenty seven different countries that compose our sample. Given the combination of this two levels of data, first of all we have run a null model without country-level or individual-level predictors which allows us to decompose the total variance in our dependent variables between individual and country levels. This permits us to estimate, through the intra-class correlation, how much of the total variance in the confidence in the parliament and the government can be explained by country-level differences only. From the value of intra-class correlation we can infer that country-level variance ranges between the 18% (CONFGOV) and the 20% (CONFPARL) of the total variance. In other words, around the 20% of the difference in the level of confidence in the parliament and around the 18% of the level of confidence in the government can be explained by the fact that a respondent lives in one country instead of another. These results lead us to employ statistical techniques that take into account the hierarchical nature of our data. Then, we have run the so-called Hausman test to compare the performance of the fixed and the random effect models. The result suggests the use of random effects models. Finally, we have computed consistent standard errors that are robust to cross-sectional heteroskedasticity of the data distribution. Table 2 here Table 2 shows the results of six regression models. The first three models test the effects of corruption, measured with CEB, CPI and CC, respectively, on the confidence in the parliament, while Models 4, 5 and 6 test the effects of the same three indicators of corruption on the confidence in the government. All the six models include both the country-level and the individual-level control variables described in the previous section. As we have hypothesized, the capacity of political systems to control corruption positively and significantly affects the citizens confidence both in the parliament and the government. When political institutions reduce the spread of corruption among elected politicians and public officials, by limiting their discretional power, citizens appear more satisfied of the political systems showing higher levels of confidence in their key political institutions. What is particularly interesting for the purpose of this research is that in the six models tested the trust-eroding impact of corruption is always significant, irrespective of whether we measure corruption with perceptions of the general public or with data gathered from experts surveys. In other words, in assessing the effects of corruption on institutional confidence to adopt a bottom-up (CEB) or a top-down (CPI and CC) approach produces the same result: the higher the level of perceived political corruption, the lower the level of confidence in the parliament and the government. Among the country-level factors tested only NEWDEM is significantly related with both the dependent variables. As expected, in young democracies of Eastern Europe, where electoral institutions are a relatively recent introduction and a stable pattern of inter-party competition is not 12

13 affirmed yet, citizens are more distrusted toward political institutions than in more consolidated democratic systems. The fact that the significant effect of corruption on institutional confidence survives to the control for the impact of new democracies is particularly relevant. In fact, given the high variability in our sample according to the level of perceived corruption (both of citizens and experts) one could presume that the trust-eroding effect of corruption was drawn by those Eastern and some Southern European countries that present high levels of corruption compared to the very low levels characterizing Scandinavian countries. Although the direction of the relationship between MAJ and the dependent variables is the expected one, its regression coefficient is never statistically significant in any of the models tested. The institutional framework (consensual Vs majoritarian systems) does not affect the citizens confidence in the parliament and the government. This basically means that whether the power is concentrated in the hands of the government or is dispersed among different institutions does not play a significant role in explaining political support in European countries. Although economic performance is considered one of the most relevant factors that influences citizens support for public institutions, our results show that neither the percapita income (loggdp) nor the economic growth (GDPG) have a significant effect on our dependent variables. Although in 2009 the international economic crisis has already began to strike many European countries, a better country s general economic situation does not influence citizens confidence in the parliament and the government. Nevertheless, at the individual-level several control variables display a significant effect on institutional confidence. Even though the country s economic performance does not affect citizens confidence in the parliament and the government, the household future economic situation (FINEXP), the life satisfaction (LIFESAT) and the social level (SOCLEV) play a significant role. In fact, citizens who expect that their household financial situation will be more prosperous in the next twelve months, are more satisfied for their life in general and place themselves in a higher level in the society are more trustful toward the parliament and the government than people who live in harder conditions. A possible explanation of this different results between country-level and individual-level economic factors is given by Banducci et al. (2003) and Hobolt and Leblond (2009) which argued that support for political institutions is determined by both socio-tropic and egocentric motives. Although on one hand citizens are concerned with the economic situation of their country, on the other they care more about their personal socio-economic conditions. Furthermore, Banducci et al. (2009) posited that the actual economic reality as summarized in official economic statistics does not necessarily agree with the economic situation perceived by different citizens. This is a potential explanation of why personal situation generates higher effects on citizens institutional confidence than general countries economic performance. Finally, while gender and education do not affect political support, the age appears as a good explanatory factor: the more citizens reach (chronologically) the maturity, the more they display confidence in the parliament and the government. Conclusions If we look at the studies that analyze the relationship between corruption and political support in a longitudinal view, we can see that this relationship has changed sign over time. During the 1960s several studies argued that corruption can boost economic and political development increasing the legitimacy of the newly established political regimes and the confidence that citizens put in institutions (Leff 1964; Bayley 1967; Huntington 1968). However, more recent studies have 13

14 empirically demonstrated that corruption is detrimental for political and economic performance (Mauro 1995; Kaufmann et al. 1999). When corruption is rooted in the society citizens distrust political institutions for their poor performance and the inefficiency in constraining corrupt practices of public servants. Various scholars have shown the negative impact of corruption on political support in Western consolidated democracies (Della Porta 2000; Anderson and Tverdova 2003) as well as in Asian (Chang and Chu 2006), African (Cho and Kirwin 2007) and Latin American countries (Seligson 2002). In line with these studies, our results confirm that corruption negatively affects the citizens confidence in political institutions also among European Union member states. In a period in which international economic and financial crisis is challenging the already tenuous relationship between rulers and ruled political corruption maintains a significant negative impact on the confidence that European Union citizens put in two key institutions of contemporary representative democracies, the government and the parliament. As expected, the higher the level of perceived corruption, the lower the citizens confidence in the parliament and the government. Much more interestingly, this relationship holds irrespective of whether we measure corruption with the most commonly used indexes gathered from experts surveys or aggregating the individual perceptions of the general public. Moreover, the effects of corruption survive for the control of a series of alternative explanatory factors both at the aggregate and at the individual level. This study shows that, at least in EU member states, the perceptions of the general public and the experts tend to move together. This disconfirms at the same time two different critiques commonly moved against measures based upon these different types of perceptions. On one side we can see that the perceptions of the experts the economic elite and the international private sector are not disconnected from the sentiments of the common citizens. On the other side, the present study shows that the general public has a quite reliable knowledge of the extent of corruption in their country and their opinion are not avulsed from the reality. Nevertheless, the indicator of corruption developed based upon citizens perceptions presents important differences with the two indexes gathered from experts surveys. Although all the three indexes of corruption depict a high variability among the sample countries, CEB places Scandinavian and Northern continental European countries clearly distant from all the other political systems. Furthermore, looking at the citizens perceptions we can recognize two distinct tendencies in act. Compared to experts evaluations citizens of some countries, especially Cyprus, Ireland and Portugal, tend to overestimate the level of corruption present in their countries, whereas citizens of other countries, such as Estonia, Poland, Slovakia and above all Italy, tend to under-estimate the situation of their countries. This last points indicate that the general public often makes in-context assessments of the extent of corruption into the public sector that risk to invalidate cross-country comparability. However, we think that a closer look and a comparison between the perceptions of small groups of experts and the general public can lead to a more comprehensive understanding of the diffusion of corruption in consolidated and recently established democratic regimes. Even though surveys on large samples of citizens are more costly and then more difficult to produce we think that further research would highly benefit from them. 14